Is there any performance improvement than the original js version?

[rubick24]

I noticed the splashy example is just calling js WebGL APIs, is there a way to render with OpenGL APIs and apply the rendered framebuffer data to canvas? I think that would have better performance.

[thomas-hoer]

I wrote a plain js port of the bouncy example.
For 1000 dots, no lines, no dashes i got on my Win 10, Intel Atom @1,6 GHz, Chrome 89 a frame rate of ~25 fps compared to ~3,7 fps for the wasm example. Also, I recompiled the go code with the go 16.2 compiler and got a frame rate of ~5 fps on the same machnine. This shows me, that the go team is still working to reduce the overhead.
index.zip

[justinclift]

@thomas-hoer As a data point, when I visit the online wasm demo page for bouncy it tells me I'm getting 60 FPS (probably capped by refresh rate, but that's just a guess).

That's on my OpenSUSE Leap 15.2 laptop, with Firefox v87. The laptop has a Ryzen 2700U cpu, and Radeon RX 560X gpu.

Not sure why your machine is giving such a low frame rate. 😦

[thomas-hoer]

Intel Atom E3950 is a low budged processor. You can use it to observe bottlenecks. You might wanna enable lines and dashed to stress your system under 60 fps and compare it with the plain js implementation.

Also I recognized that the WebAssembly compiled with Go 1.16 was about 20% faster than your binaries from Go1.13.